Agent bio bug bounty call
OpenAI puts $25K on the table. Find the jailbreak in ChatGPT's agent mode.
Why it matters
OpenAI is crowdsourcing adversarial testing of its agentic systems before wider deployment, signaling both confidence in safety architecture and acknowledgment that internal testing alone is insufficient—a bellwether for how AI labs will approach agent safety governance.
The key facts
10 to knowOpenAI launches Bio Bug Bounty program for ChatGPT agent safety testing
Bounty cap: $25,000 per vulnerability
Focus: Universal jailbreak prompts and agent-specific safety gaps
Program targets researcher community for adversarial validation
Timing: Mid-July 2025, suggesting agent deployment phase
OpenAI launched Bio Bug Bounty program for ChatGPT agents
Up to $25,000 reward per valid vulnerability
Focus: universal jailbreak prompts and agent safety
Open to external researchers
Tests agent behavior under adversarial conditions
Go to the source
OpenAI Blogopenai.com
Publisher excerpt: OpenAI invites researchers to its Bio Bug Bounty. Test the ChatGPT agent’s safety with a universal jailbreak prompt and win up to $25,000.